The white-chinned petrel is a subantarctic seabird that requires urgent implementation of conservation measures for the species. At sea, adults suffer heavy mortality due to fisheries' practices. On land, introduced rats prey on chicks at several localities, and we test here if and how rats can be efficiently controlled. Since 1994, we have conducted an intensive ratcontrol program during each breeding season in a white-chinned petrel colony on Ile de la Possession (150 km2; Crozet archipelago, southern Indian Ocean), which had been monitored since 1986. On the same island, a control white-chinned petrel colony, where no poisoning occurred, was also monitored, and we assessed the seasonal variations of rat abundance. We compared three situations: high rat-poisoning, low ratpoisoning and control conditions without poisoning. Low-poisoning trials performed in our experimental colony between 1988 and 1991 did not lead to higher chick production than for the previous two control years. However, petrel-breeding success was significantly higher when intensive poisoning occurred (50%) than for the previous years (16%). The duration of our study (8 years before intensive poisoning, plus 8 years afterwards), combined with a comparison of petrel annual breeding success between our experimental and control colonies, allowed us to assess more effectively the impact of rats. Forty-one per cent of breeding failures occurring in non-poisoned areas were attributed to rats. We conclude that threatened insular bird populations can be conserved and restored in localities even where total rat eradication is not possible. However, only intensive and repeated (long-term) poisoning will control rats sufficiently.